Anthropic's research team has achieved a breakthrough in understanding the internal architecture of large language models. During an analysis of the Claude family of models, a structure was identified that developers did not intentionally design. This refers to the so-called "J-space" — an internal mechanism that functions as a shared workspace accessed by various components of the model.
According to published data (research from July 6), J-space is a virtual "whiteboard" where key information is collected and transmitted across the entire model. When Claude answers a question, solves a problem, or follows an instruction, critically important data appears in this space so that different parts of the model can work with it. Researchers used a tool called "J-lens," which allows real-time observation of information movement within the model. Surprisingly, J-space emerged spontaneously during the training process — it was not directly built in by the designers.
This concept remarkably echoes what neuroscience calls the "global workspace." In humans, this system provides access to important information for multiple thought processes simultaneously. For example, when a person hears a question, recalls a relevant fact, and decides how to answer, the brain consolidates all necessary data in one place. Claude operates in a similar way: the model can describe the contents of J-space upon request and even modify it if asked. Moreover, when researchers manually altered J-space, Claude's answers and its behavior when performing tasks changed accordingly.
Why is this important for AI safety and interpretability?
This discovery has enormous significance for artificial intelligence safety. If scientists can track J-space activity, they will have a chance to identify hidden motives in the behavior of AI models. This will make it more effective to detect moments when the system begins to operate unreliably, including during attacks like prompt injection (inserting malicious instructions into a query). Even partial access to such a "conscious" processing layer opens up important prospects for research.
For now, capabilities are limited: most of Claude's information processing still occurs outside J-space. Nevertheless, Anthropic has already open-sourced the implementation of J-lens and posted a demo on Neuronpedia, inviting the research community to verify the results in practice.
This research builds on a series of earlier works. In October 2025, the company published a report on emerging introspective awareness, and in April 2025, it launched initiatives to study model well-being. An important caveat: Anthropic emphasizes that it is not claiming Claude possesses consciousness or subjective experience. The study uses the term "consciously accessible" information, borrowed from cognitive science — it does not imply the presence of actual consciousness.
Expert opinion: The discovery of J-space is not just an academic curiosity, but a step toward turning "black boxes" into "white boxes." For the crypto industry, where AI is increasingly used in market analysis and asset management, the ability to look inside a model's "head" and understand its motives is a matter of trust and security. Monitoring J-space could become a standard for verifying AI agents in DeFi.